yuanbw2025/storyforge

What it solves

StoryForge is an open-source, local-first AI narrative creation platform designed to handle the complexities of long-form storytelling. It addresses the challenge of maintaining consistency across massive works (targeting up to a million words) by breaking the creative process into structured steps and separating the "world engine" (semantic content) from the actual narrative output.

How it works

The system uses a "Harness" architecture to manage AI agents. Instead of direct generation, it follows a strict pipeline: the agent receives a registered context, generates a candidate output (CreativeArtifact), which is then validated against a schema and presented to the author for approval or editing before being officially adopted into the project.

To maintain long-term consistency, it employs a "Context Gateway" and three single sources of truth: a context assembler for AI reading, a field registry for AI writing, and project tables for data governance. Data is stored locally in the browser's IndexedDB or local files to ensure privacy.

Who it’s for

• Novelists and writers who need AI assistance for long-form narratives, world-building, and detailed outlining.
• Creators looking to convert novels into scripts or manga.
• Game masters and players interested in TTRPGs, character chats, and text-based games based on frozen world versions.

Highlights

• Multi-format support: Dedicated workflows for long-form novels, short stories, scripts, and manga.
• Local-first privacy: Stores manuscripts and project data in IndexedDB, requiring users to provide their own API keys.
• Structured creation: A step-by-step pipeline from world-building and character design to scene outlines and final text generation.
• Versioned World Engine: A system that saves semantic content and versions, allowing read-only frozen versions to be used for interactive experiences like TTRPGs.